Security Guards

33-9032.00
Median wage $38,020/yr1,283,470 employed (US)Rank #569 of 923 scored · top 62% by substitution

Guard, patrol, or monitor premises to prevent theft, violence, or infractions of rules. May operate x-ray and metal detector equipment.

Sub-scores

0–100 · band = confidence interval from rater disagreement

Substitution24
Exposure21
Augmentation52

Substitution — the headline: capability discounted by cost, barriers and adoption.

Exposure — technical capability alone, regardless of whether anyone deploys it.

Augmentation — how much AI assists without replacing. High here + moderate substitution = a changing job, not a disappearing one.

Tasks on the substitution scale

14 rated tasks, binned by substitution score.

Position among all scored occupations

Distribution of 923 occupation scores; the marker is this occupation.

Tasks with substitution ≥ 70

0%

Run 1.0.0-draft.1 · computed 2026-08-05 · rater panel: claude-sonnet-5, claude-haiku-4-5-20251001 · intervals span rater disagreement.

Why this score

The five weighted dimensions of the composite, averaged across this occupation's tasks (importance-weighted, panel mean). Exact weights and formulas: /api/v1/methodology.

Task automatabilityw 35%20

panel mean rating 1.8/5 → substitution pressure 20/100

Technical feasibility todayw 20%22

panel mean rating 1.9/5 → substitution pressure 22/100

Cost vs. human wagew 15%27

panel mean rating 2.1/5 → substitution pressure 27/100

Adoption barriersw 20%inverted — strong barriers lower the score33

panel mean rating 3.7/5 (barrier strength) → substitution pressure 33/100

Sector adoption velocityw 10%22

panel mean rating 1.9/5 → substitution pressure 22/100

Task breakdown (14 tasks)

Substitution pressure per task, weighted by O*NET importance in the composite. Expand a task for the full rater audit trail — every rating, every model, every rationale.

Answer telephone calls to take messages, answer questions, and provide information during non-business hours or when switchboard is closed.

67

CI 56–79 · exposure 62 · augmentation 63 · importance 4.5/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5Information-sector and large enterprise adoption of automated call systems is mature and widespread. Most large organizations already route after-hours calls through automated systems, indicating fast, deep adoption in digitized sectors.
Sector adoption velocityclaude-sonnet-52/5Security services is a lower-digitization sector with slower AI adoption compared to information/finance sectors, though call-answering automation is spreading via facilities-management vendors.
Augmentation potentialclaude-haiku-4-5-202510013/5AI can assist human security staff by pre-screening calls, summarizing messages, and suggesting responses, raising their efficiency. However, the task is already quite automatable end-to-end, so augmentation's independent value is moderate.
Augmentation potentialclaude-sonnet-54/5AI call-handling and transcription tools can significantly reduce guard workload for routine messages while escalating urgent calls, improving overall responsiveness.
Task automatabilityclaude-haiku-4-5-202510014/5AI systems can reliably handle most incoming calls by understanding questions, providing factual information, and recording messages with high accuracy. However, edge cases requiring nuanced judgment or escalation to human staff prevent a full 5 rating, though the core task (answering, routing, recording) easily meets the 50% time-saving threshold.
Task automatabilityclaude-sonnet-53/5AI phone/voice agents can already answer calls, take messages, and provide basic information reliably, but part of the task (situational judgment during security incidents) still needs human oversight.
Adoption barriersclaude-haiku-4-5-202510012/5There are minimal regulatory or licensing barriers to automating after-hours call answering; no law requires a licensed human to answer phones. The main friction is organizational inertia and customer preference for human contact in some contexts, which is surmountable.
Adoption barriersclaude-sonnet-52/5No licensing requirement specifically for phone answering, though some employers may prefer a human security presence for liability/emergency response reasons.
Cost vs. human wageclaude-haiku-4-5-202510015/5Cloud-based automated call systems cost pennies per call compared to fully-loaded human security guard wages (typically $15–25/hour). The cost advantage is well over an order of magnitude.
Cost vs. human wageclaude-sonnet-54/5Automated answering/voice-bot services cost a small fraction of paying a guard's wage for after-hours phone coverage, though setup and monitoring add some overhead.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed IVR and AI call-handling systems (e.g., voice assistants, commercial call-routing platforms) already perform this task in production at many organizations. Minor limitations around accent handling and complex multi-step queries prevent a 5, but production reliability is demonstrated.
Technical feasibility todayclaude-sonnet-53/5Virtual receptionist and IVR/voice-AI products are deployed widely for message-taking and FAQ answering, but integration with security-specific escalation protocols is narrower and less proven.

Write reports of daily activities and irregularities, such as equipment or property damage, theft, presence of unauthorized persons, or unusual occurrences.

54

CI 51–56 · exposure 50 · augmentation 75 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Security is a fragmented sector with many small operators and legacy processes; while large enterprises pilot AI report systems, production adoption remains limited and rollout is slow compared to digitized white-collar functions.
Sector adoption velocityclaude-sonnet-52/5Security services is a lower-digitization, often small-firm sector with slow, uneven AI tool adoption despite some incident-reporting software integrating AI features.
Augmentation potentialclaude-haiku-4-5-202510014/5AI can substantially assist by auto-drafting incident summaries from guard notes or sensor feeds, flagging patterns, and suggesting report structure, enabling faster and more complete documentation while the guard retains final oversight and judgment.
Augmentation potentialclaude-sonnet-54/5AI-assisted dictation and drafting tools meaningfully speed up report writing and improve clarity while the guard remains responsible for accuracy and content.
Task automatabilityclaude-haiku-4-5-202510013/5Current AI can extract and organize incident details from guard observations or sensor logs and generate structured reports with significant time savings, but requires human judgment to interpret context, assess threat significance, and determine what constitutes an 'irregularity' worthy of escalation.
Task automatabilityclaude-sonnet-53/5AI can draft or structure incident reports from dictated notes or bullet points, but the underlying observation, judgment about what's 'unusual,' and factual accuracy still require a human guard on site.dictated input.It's the write-up portion that's automatable, not the whole task.
Adoption barriersclaude-haiku-4-5-202510013/5Reports may feed into liability, insurance, or regulatory compliance chains; organizations often require human sign-off and maintain preference for human-authored documentation to defend incident response decisions in disputes.
Adoption barriersclaude-sonnet-52/5Reports may be used as evidence in legal/insurance matters, creating some liability concern, but no licensing requirement mandates a human author the report itself.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI-powered report generation (NLP + templates) costs far less per report than paying a security officer's loaded wages for the same writing task, though integration and oversight overhead moderates the advantage.
Cost vs. human wageclaude-sonnet-54/5Transcription and drafting via AI tools costs a fraction of the guard's time spent writing, though some human oversight cost remains.
Technical feasibility todayclaude-haiku-4-5-202510013/5Production systems exist (incident report generation, log analysis tools) but often require substantial human input, template selection, and review; error rates in categorizing incidents or missing contextual details remain material in real deployments.
Technical feasibility todayclaude-sonnet-53/5Voice-to-text and generative writing tools are widely deployed and used in report drafting, but most guard services still require human review/editing for accuracy and legal defensibility, limiting reliability at scale.

Operate detecting devices to screen individuals and prevent passage of prohibited articles into restricted areas.

43

CI 16–70 · exposure 42 · augmentation 75 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510014/5High-traffic security checkpoints (airports, borders, corporate campuses) have adopted automated screening at scale; many facilities now deploy AI-assisted or fully automated detection alongside or replacing initial-line human screening.
Sector adoption velocityclaude-sonnet-52/5Physical security is a low-digitization, in-person sector; while some AI-enhanced scanning tools are piloted in aviation/critical infrastructure, broad deployment replacing guard-operated screening is slow.
Augmentation potentialclaude-haiku-4-5-202510015/5AI systems substantially augment guard productivity by flagging suspicious items, handling routine checks, and reducing manual inspection workload, allowing guards to focus on higher-judgment decisions and response—a clear productivity multiplier.
Augmentation potentialclaude-sonnet-53/5AI-enhanced image recognition can help guards more quickly and accurately flag suspicious items in x-ray/scanner feeds, improving detection speed and accuracy while the guard remains responsible for decisions and physical response.
Task automatabilityclaude-haiku-4-5-202510014/5AI-driven screening systems (metal detectors, X-ray image recognition, thermal imaging) can already automate most detection and flagging functions with high accuracy. However, final discretionary decisions on borderline cases and physical intervention still require human judgment, preventing a full 5.
Task automatabilityclaude-sonnet-51/5This requires physical presence to operate detection equipment (metal detectors, x-ray scanners), engage with individuals passing through, and physically intervene when contraband is found—none of which current AI systems can perform end-to-end without embodiment.
Adoption barriersclaude-haiku-4-5-202510013/5Regulatory frameworks often mandate human presence and liability for errors at checkpoints, and organizational policies typically require a guard to authorize blocking or escalation. These requirements create meaningful friction but do not absolutely prohibit automation in all contexts.
Adoption barriersclaude-sonnet-54/5Security/screening in restricted areas is often governed by regulations, security clearances, and liability concerns that require human authorization and hands-on physical enforcement, creating strong barriers to full replacement.
Cost vs. human wageclaude-haiku-4-5-202510014/5AI screening hardware and software amortized across high-traffic checkpoints cost far less per screening event than the fully loaded wage of a security guard monitoring manually. Initial capital investment is significant but operational cost per check is low.
Cost vs. human wageclaude-sonnet-52/5AI-assisted detection hardware exists but is capital-intensive and still requires human staffing for physical enforcement, so overall cost per screening point doesn't dramatically undercut wage costs when a human presence is still mandated.
Technical feasibility todayclaude-haiku-4-5-202510014/5Deployed AI systems for baggage screening, threat detection, and anomaly identification exist in production at airports, border crossings, and high-security facilities. Performance is reliable for clear-cut cases, though edge cases and adversarial scenarios require human confirmation.
Technical feasibility todayclaude-sonnet-52/5Automated scanning tech (AI-assisted x-ray image analysis, computer vision at checkpoints) exists in some airports and high-security facilities, but a human guard operating the device, making judgment calls, and physically stopping violators remains standard practice.

Monitor and adjust controls that regulate building systems, such as air conditioning, furnace, or boiler.

30

CI 25–35 · exposure 25 · augmentation 50 · importance 2.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While large enterprises deploy advanced BMS, most security guard-staffed facilities use basic monitoring; adoption of autonomous control adjustment remains slow due to regulatory requirements, risk aversion, and the need for integration with legacy infrastructure.
Sector adoption velocityclaude-sonnet-52/5Security services and facilities management are historically slow-adopting sectors for AI-driven automation; smart building tech is spreading but security guard roles specifically remain largely unautomated in practice.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-assisted monitoring (alerts, trend analysis, predictive maintenance recommendations) can meaningfully help a human operator make faster, better-informed decisions about control adjustments, though the final decision and action typically remain with the human.
Augmentation potentialclaude-sonnet-53/5AI-enabled building management dashboards and anomaly alerts can meaningfully assist guards in monitoring HVAC/boiler systems, flagging issues faster than manual checks, though adjustments and final judgment remain human-driven.
Task automatabilityclaude-haiku-4-5-202510012/5Only limited parts of monitoring building systems can be automated today (sensor data collection, basic alerts); adjusting controls requires real-time environmental sensing, decision-making under uncertainty, and physical intervention that current AI systems cannot reliably perform end-to-end without human oversight.
Task automatabilityclaude-sonnet-52/5Physical presence and control adjustment via building management systems can be partially automated with smart BMS/IoT sensors, but this task as described requires a human present to monitor and physically adjust equipment, limiting full automation.rate
Adoption barriersclaude-haiku-4-5-202510014/5Building codes and safety regulations typically mandate that HVAC system adjustments be performed or approved by licensed technicians; liability for system failure (temperature excursions, equipment damage) creates strong friction against full automation without credentialed human sign-off.
Adoption barriersclaude-sonnet-52/5No licensing requirement specifically ties this task to a human, though safety/liability concerns around building systems (fire, gas, electrical) create some friction against full unsupervised automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Integration of AI-enabled BMS monitoring and control requires significant setup, specialized HVAC knowledge, system calibration, and ongoing human oversight; total cost often exceeds the wage savings from reducing one security guard's time on this task.
Cost vs. human wageclaude-sonnet-52/5While automated BAS sensors are cheap to run per data point, replacing the guard's broader monitoring/adjustment role entirely still requires human oversight and physical presence, keeping all-in costs closer to comparable rather than order-of-magnitude cheaper.
Technical feasibility todayclaude-haiku-4-5-202510012/5Building management systems (BMS) exist for monitoring and can log data, but fully autonomous control adjustment without human supervision remains rare in production; most deployed systems require human operators to interpret alerts and manually adjust controls for safety and liability reasons.
Technical feasibility todayclaude-sonnet-52/5Building automation systems (BAS) exist and are deployed widely for HVAC monitoring, but they typically supplement rather than replace the guard's monitoring role, and adjustment often still requires human judgment or intervention on-site.

Monitor and authorize entrance and departure of employees, visitors, and other persons to guard against theft and maintain security of premises.

28

CI 25–30 · exposure 25 · augmentation 63 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption is slow outside large, highly digitized facilities (data centers, tech campuses). Most enterprises, small businesses, and physical locations continue to rely on human guards, and sectors are cautious about delegating access authorization fully to automation given security liability.
Sector adoption velocityclaude-sonnet-52/5Security services is a moderately digitized but physically-oriented, often small-business-dominated sector; AI camera analytics are being piloted but full autonomous guarding remains rare in production.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-powered cameras, anomaly detection, and real-time alerts can assist guards by highlighting suspicious activity and automating routine logging, meaningfully raising guard efficiency. However, the core judgment task remains human-driven, so augmentation is useful but not transformative.
Augmentation potentialclaude-sonnet-54/5AI-powered video analytics, anomaly detection, and access logs significantly help guards monitor more entrances and flag threats faster, meaningfully boosting situational awareness while the guard remains responsible for decisions.
Task automatabilityclaude-haiku-4-5-202510012/5While camera monitoring and access control systems can log and flag unusual activity, the task requires real-time judgment about who should be authorized, intent assessment, and adaptive decision-making in dynamic situations. AI cannot reliably replace the human discretion needed to evaluate suspicious behavior or make judgment calls on edge cases, so meaningful automation falls well short of 50% time savings at equal quality.
Task automatabilityclaude-sonnet-52/5Access-control components (badge scanning, camera-based ID checks) can be automated, but real-time judgment about suspicious behavior, physical intervention, and handling exceptions still requires a human presence on-site.atability is partial at best.
Adoption barriersclaude-haiku-4-5-202510014/5Significant barriers exist: organizational liability falls on whoever authorizes access, premises security is often legally tied to accountable personnel, and many facilities face insurance or regulatory requirements for human presence at entry points. Customer confidence and liability asymmetry strongly discourage full replacement.
Adoption barriersclaude-sonnet-53/5No licensing barrier for the monitoring task itself, but liability for security breaches, need for physical intervention capability, and customer/insurance expectations of human presence create moderate friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5While camera and access-control infrastructure has modest per-transaction costs, the need for human oversight and intervention for authorization decisions means total cost (infrastructure + oversight) remains comparable to or exceeds a security guard's loaded wage for the same output of vetted access control.
Cost vs. human wageclaude-sonnet-52/5Camera and badge systems have upfront and maintenance costs comparable to guard wages when factoring in monitoring, false-alarm response, and system upkeep; full replacement of a guard's judgment and physical deterrence is not yet cheaper.
Technical feasibility todayclaude-haiku-4-5-202510012/5Computer vision systems exist for monitoring premises and some access control can be automated (badge readers, turnstiles), but they cannot independently authorize entrance/departure with the human-like judgment required to prevent sophisticated theft or security breaches. Deployed systems assist monitoring rather than performing the authorization task end-to-end reliably.
Technical feasibility todayclaude-sonnet-52/5Automated access control systems, turnstiles, and AI-based video analytics are deployed widely, but they typically supplement rather than replace a guard's physical presence and judgment calls, especially for ambiguous situations.

Inspect and adjust security systems, equipment, or machinery to ensure operational use and to detect evidence of tampering.

28

CI 25–30 · exposure 25 · augmentation 50 · importance 4.2/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5While security monitoring is digitizing, actual replacement of inspection and adjustment tasks remains minimal; most organizations use AI as supplementary alerting, not primary task execution.
Sector adoption velocityclaude-sonnet-52/5Security services are a lower-digitization, physical-labor-heavy sector where AI adoption for hands-on equipment inspection remains in early pilot stages rather than widespread production use.
Augmentation potentialclaude-haiku-4-5-202510013/5Automated alerts and log analysis can assist guards in prioritizing patrols and identifying anomalies, improving efficiency in routine monitoring without removing human responsibility for final judgment and remediation.
Augmentation potentialclaude-sonnet-53/5AI-powered sensors, alerts, and analytics can flag anomalies or potential tampering, helping guards prioritize where physical inspection is most needed, though the core physical task still requires the human.
Task automatabilityclaude-haiku-4-5-202510012/5Routine monitoring and log review could be partially automated, but the task requires physical inspection, contextual judgment about tampering evidence, and adjustment of complex systems—tasks current AI struggles with without significant human intervention.
Task automatabilityclaude-sonnet-52/5Visual inspection, physical adjustment, and tampering detection require on-site physical presence and manual dexterity that current AI systems cannot replicate end-to-end.'
Adoption barriersclaude-haiku-4-5-202510014/5Security operations are heavily regulated, often require licensed security personnel to conduct inspections, and carry legal liability if tampering is missed; organizational protocols typically mandate human sign-off on system integrity.
Adoption barriersclaude-sonnet-53/5No strict licensing requirement mandates a human specifically for this task, but liability for missed tampering and the need for physical access create meaningful operational friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5Camera systems and automated monitoring are costly to install and maintain, and still require security personnel oversight; total cost approaches or exceeds the wage of regular inspection patrols.
Cost vs. human wageclaude-sonnet-52/5AI monitoring tools add sensor and software costs on top of still-needed human labor for physical inspection and adjustment, so total cost is not clearly cheaper than a guard performing the full task.
Technical feasibility todayclaude-haiku-4-5-202510012/5Camera-based monitoring and alert systems exist, but they have high false-positive rates and cannot reliably detect tampering across diverse equipment types or perform physical adjustments; deployed products require substantial human verification.
Technical feasibility todayclaude-sonnet-52/5Some AI-enabled sensors and anomaly-detection software exist for monitoring systems remotely, but physical inspection and hands-on adjustment of equipment remain unaddressed by deployed products.

Call police or fire departments in cases of emergency, such as fire or presence of unauthorized persons.

25

CI 20–30 · exposure 25 · augmentation 63 · importance 4.4/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Despite digitization of security, actual AI-driven emergency calling remains minimal; most deployments still require a human to verify and place the call, reflecting both regulatory friction and organizational risk aversion.
Sector adoption velocityclaude-sonnet-52/5Physical security is a low-digitization sector with slow, uneven adoption of automated alerting/monitoring technology, mostly in higher-end commercial properties.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly assists by automatically detecting anomalies (fires, intruders) via sensors and alerting the guard, enabling faster human decision-making and call placement while the guard retains control and judgment of whether to contact emergency services.
Augmentation potentialclaude-sonnet-53/5AI-based surveillance and anomaly detection tools can alert guards to potential emergencies faster, improving response time while the guard retains decision authority to call authorities.
Task automatabilityclaude-haiku-4-5-202510012/5AI systems can classify some emergency scenarios from sensor data or reports, but cannot reliably determine the nuanced judgment of what constitutes a genuine emergency versus a false alarm, nor can they legally place official emergency calls in most jurisdictions without human authorization and verification.
Task automatabilityclaude-sonnet-52/5This requires real-time physical presence, sensory judgment about threats, and split-second decision-making that current AI cannot reliably replicate end-to-end, though alert-triggering systems can flag some events.
Adoption barriersclaude-haiku-4-5-202510015/5Strong legal and regulatory barriers exist: emergency services require verified human callers, dispatch systems are designed around human contact, and liability for false alarms or missed emergencies falls on the organization making the call, effectively requiring human sign-off.
Adoption barriersclaude-sonnet-53/5No strict licensing requirement mandates a human specifically make this call, but liability for false alarms, need for situational judgment, and organizational reliance on human verification create meaningful friction.
Cost vs. human wageclaude-haiku-4-5-202510012/5The infrastructure to automate emergency call-making (sensors, verification systems, monitoring) is expensive, and the cost of false dispatches is severe; total cost of ownership is likely comparable to or exceeds the human task cost.
Cost vs. human wageclaude-sonnet-52/5AI monitoring systems have upfront and maintenance costs comparable to or sometimes exceeding a guard's marginal wage for this specific sub-task, especially when false alarms incur police response costs.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI can assist with emergency detection through cameras or alerts, no deployed product reliably makes independent emergency calls to police/fire; this remains a human-verified function in practice due to liability and legal requirements.
Technical feasibility todayclaude-sonnet-52/5Deployed smart security systems (cameras with AI detection, alarm monitoring services) can trigger alerts, but human verification and judgment before contacting emergency services remains standard practice due to false-positive risks.

Patrol industrial or commercial premises to prevent and detect signs of intrusion and ensure security of doors, windows, and gates.

24

CI 16–32 · exposure 17 · augmentation 63 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510013/5Organizations are adopting AI-enhanced surveillance and monitoring tools, but human security guards remain the norm in most industrial and commercial settings; automation is augmentative rather than replacement-focused in production environments.
Sector adoption velocityclaude-sonnet-52/5Security services is a traditionally low-digitization sector; adoption of camera/robot patrol supplements is growing but production-scale replacement of human patrols remains rare.
Augmentation potentialclaude-haiku-4-5-202510014/5AI-powered monitoring systems significantly enhance security guards' productivity by automatically flagging suspicious activity, tracking door/window status, and centralizing alerts, allowing guards to focus on high-risk areas and respond faster to real threats while the human remains the final decision-maker.
Augmentation potentialclaude-sonnet-53/5AI-enabled cameras, motion sensors, and analytics can alert guards to anomalies and prioritize patrol routes, meaningfully assisting but not transforming the core physical task.
Task automatabilityclaude-haiku-4-5-202510012/5While AI-powered cameras and sensors can detect some intrusions and monitor entry points, the task requires physical presence, real-time decision-making in unpredictable situations, and human judgment about suspicious behavior—capabilities that current autonomous systems cannot reliably provide end-to-end without substantial human oversight.
Task automatabilityclaude-sonnet-51/5Physical patrolling requiring mobility, on-site presence, and real-time physical intervention cannot be performed end-to-end by current AI systems; sensors and cameras can supplement but not replace the physical patrol.20
Adoption barriersclaude-haiku-4-5-202510014/5Liability and legal responsibility for security breaches typically fall on property owners, who face insurance and regulatory requirements that often mandate human security presence; customer expectations and legal frameworks generally require a licensed or accountable human to sign off on security protocols.
Adoption barriersclaude-sonnet-53/5No licensing strictly requires a human patroller, but liability for security breaches, insurance requirements, and client preference for human presence create moderate friction against pure automation.
Cost vs. human wageclaude-haiku-4-5-202510012/5Security camera and sensor systems are capital-intensive and require ongoing maintenance, monitoring, and human security staff for verification and response, making the all-in cost comparable to or higher than hiring human guards, especially for 24/7 coverage.
Cost vs. human wageclaude-sonnet-52/5Robotic patrol systems, sensor networks, and monitoring software have high upfront capital and maintenance costs that often exceed or match guard wages when factoring installation, coverage gaps, and human oversight needed.
Technical feasibility todayclaude-haiku-4-5-202510012/5Deployed security camera systems with AI detection exist and can flag motion or open doors, but they produce high false-positive rates and cannot physically patrol or respond to intrusions; real-world security still requires human guards to validate alerts and handle contingencies.
Technical feasibility todayclaude-sonnet-52/5Deployed products like camera-based monitoring, drones, and robotic patrol units exist in some facilities but are narrow-scope supplements rather than reliable full replacements for physical guard patrols.

Answer alarms and investigate disturbances.

15

CI 5–25 · exposure 13 · augmentation 63 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption of AI monitoring is growing in controlled facilities, but most organizations retain human security guards for alarm response and investigation. Deployment remains limited to specific sectors (data centers, some retail) and pilot programs rather than deep, rapid penetration across security industries.
Sector adoption velocityclaude-sonnet-51/5Physical security is a low-digitization, on-site sector with minimal AI-driven displacement of physical response duties.
Augmentation potentialclaude-haiku-4-5-202510014/5AI significantly assists security guards by automating alarm triage, identifying anomalies in video feeds, and prioritizing which disturbances warrant investigation, allowing guards to focus on higher-value responses. This augmentation measurably improves guard productivity and situational awareness while keeping humans in control.
Augmentation potentialclaude-sonnet-53/5AI-enabled sensors, cameras, and alarm analytics can help guards detect and prioritize disturbances faster, aiding but not replacing the physical response.
Task automatabilityclaude-haiku-4-5-202510012/5While AI can monitor alarms and flag disturbances via video/sensor analysis, physical investigation of disturbances requires on-site presence and situational judgment that current autonomous systems cannot reliably perform. Alarm response also involves human judgment about threat assessment and escalation that remains difficult to fully automate.
Task automatabilityclaude-sonnet-51/5This requires physical presence, real-time perception of an environment, and often physical intervention or judgment calls on-site, which current AI cannot perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510014/5Strong barriers exist: liability and legal responsibility for security incidents fall on the organization, not the AI; property owners often face insurance and contractual requirements for human security presence; and the need for physical on-site presence and human judgment in potentially dangerous situations creates regulatory and organizational friction.
Adoption barriersclaude-sonnet-54/5Physical response to disturbances often carries liability, safety, and sometimes licensing/authorization requirements (e.g., use of force, detainment) that require a human on-site.
Cost vs. human wageclaude-haiku-4-5-202510012/5AI-driven monitoring can reduce overhead, but the physical investigation component requires either deployed security robots (expensive and immature) or human responders. The total cost of an autonomous alarm-investigation system remains comparable to or higher than employing security guards in most settings.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute performing the physical investigation task, so cost comparison favors the human by default since the AI cannot deliver the output.
Technical feasibility todayclaude-haiku-4-5-202510012/5Video surveillance AI and alarm-monitoring systems exist in production, but they typically alert humans rather than autonomously investigate. No deployed products reliably perform the full investigation and response cycle without human security personnel on-site to assess and act on threats.
Technical feasibility todayclaude-sonnet-51/5No deployed product physically investigates disturbances; AI-based alarm systems can flag events but cannot perform the physical investigation itself.

Circulate among visitors, patrons, or employees to preserve order and protect property.

12

CI 7–16 · exposure 5 · augmentation 50 · importance 4.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510012/5Adoption is slow despite decades of surveillance technology; most physical security operations still employ on-site human patrols. Organizations are primarily augmenting (adding cameras) rather than replacing guards, and adoption remains concentrated in high-value facilities.
Sector adoption velocityclaude-sonnet-52/5Security services sector is a physical, moderately digitized industry; AI adoption is mostly in surveillance analytics and alerting, not replacing patrol duties, so displacement is slow.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-powered surveillance can alert guards to anomalies, map patron flow, and identify suspicious activity, meaningfully supporting patrol efficiency and decision-making, though the human remains essential for actual intervention and judgment.
Augmentation potentialclaude-sonnet-53/5AI-powered camera systems, anomaly detection, and communication tools can help guards prioritize attention and respond faster, offering moderate assistance to the human performing rounds.
Task automatabilityclaude-haiku-4-5-202510011/5Circulating among people to preserve order and protect property requires real-time physical presence, nuanced judgment about threats, and contextual responsiveness that current AI systems cannot perform end-to-end. AI cannot autonomously patrol physical spaces or make complex situational decisions about when to intervene.
Task automatabilityclaude-sonnet-51/5This requires physical presence, mobility, and real-time human judgment in a physical space; no off-the-shelf AI can physically patrol and intervene to preserve order today.
Adoption barriersclaude-haiku-4-5-202510014/5Legal liability for property damage or injury, insurance requirements, and the expectation of immediate human response to incidents create strong barriers. Many venues require visible human security presence for both deterrence and regulatory compliance.
Adoption barriersclaude-sonnet-54/5Security guard roles often require licensing, liability considerations, and physical authority (e.g., detaining individuals) that legally and practically require a human presence.
Cost vs. human wageclaude-haiku-4-5-202510012/5Security cameras and monitoring systems are relatively inexpensive, but they do not eliminate the need for human guards to physically circulate and respond. Total cost of an automated monitoring setup plus residual human oversight rarely undercuts a guard's loaded wage.
Cost vs. human wageclaude-sonnet-51/5Physical robotic patrol systems capable of substituting for a guard's mobility and judgment are expensive, require significant infrastructure, and are not cheaper than human labor at scale today.
Technical feasibility todayclaude-haiku-4-5-202510012/5While AI-powered cameras and sensors can detect some anomalies (motion, unauthorized access), no deployed system can reliably circulate, assess social dynamics, and make protective decisions without a human present. Surveillance assists but does not replace the patrol function.
Technical feasibility todayclaude-sonnet-51/5While camera-based surveillance and anomaly detection products exist, they do not perform physical circulation, deterrence, or hands-on order preservation; that remains research/robotics-stage for general deployment.

Lock doors and gates of entrances and exits to secure buildings.

11

CI 0–21 · exposure 0 · augmentation 25 · importance 4.9/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5This task remains almost entirely human-performed because it requires physical presence and real-world manipulation. There is minimal adoption of automation for this specific task in security operations.
Sector adoption velocityclaude-sonnet-52/5Security services adopt access-control automation (keycards, smart locks) at a moderate pace, but full replacement of guard-performed locking duties remains slow given physical security contexts.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could assist marginally by reminding guards to lock specific doors or logging lock status, but the core task of physically securing entrances requires human presence and mechanical action.
Augmentation potentialclaude-sonnet-52/5Smart lock systems and building automation can notify guards of lock status or automate some entry points, offering modest assistance, but this doesn't fundamentally transform how guards perform the physical task.
Task automatabilityclaude-haiku-4-5-202510011/5Locking physical doors and gates requires mechanical operation in the real world. Current AI systems cannot physically manipulate physical objects without specialized robotics, which are not widely deployed for this purpose.
Task automatabilityclaude-sonnet-51/5This is a physical manipulation task requiring presence at doors/gates; current AI systems cannot physically lock doors and gates.'
Adoption barriersclaude-haiku-4-5-202510015/5Physical security and liability concerns create strong barriers: building owners and insurers typically require a human presence to visually confirm locked status and respond to access issues, and negligence liability falls heavily on the operator.
Adoption barriersclaude-sonnet-52/5No licensing requirement specifically for locking doors, but physical infrastructure constraints and the need for human presence for verification/monitoring create moderate friction against pure automation.
Cost vs. human wageclaude-haiku-4-5-202510011/5The cost of deploying robots capable of physically locking doors would far exceed the cost of a security guard performing this task, making the economic case unfavorable.
Cost vs. human wageclaude-sonnet-52/5Automated electronic locks can be cheaper long-term than a guard's wage for this narrow function, but require capital investment, retrofitting, and don't replace the broader guard role, so cost comparison is mixed.
Technical feasibility todayclaude-haiku-4-5-202510011/5No mainstream product performs end-to-end physical door/gate locking autonomously in deployed security settings. Any automation would require custom robotics integration, not general-purpose AI.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs physical locking of doors/gates as a general capability; only fixed automated locking systems exist, which are not AI-driven task substitution but pre-installed mechanical/electronic controls.

Escort or drive motor vehicle to transport individuals to specified locations or to provide personal protection.

7

CI 0–14 · exposure 8 · augmentation 25 · importance 3.7/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Security firms remain labor-intensive and cautious about automation. Client preference for human judgment, perceived threat assessment, and presence is strong. Adoption of autonomous protective escort remains nearly non-existent in practice.
Sector adoption velocityclaude-sonnet-51/5Security guard services are a low-digitization, physically embedded sector with minimal AI-driven displacement in protective escort functions.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could modestly assist with route planning, vehicle diagnostics, or threat-alert flagging, but human security judgment, crisis response, and protective presence remain central to the task. Augmentation value is limited because the core protective decision-making cannot be delegated.
Augmentation potentialclaude-sonnet-52/5AI can assist with route planning, threat intelligence, or communication logistics, but offers little help with the core physical escort and protection duties.
Task automatabilityclaude-haiku-4-5-202510012/5While autonomous vehicles can handle basic driving to specified locations in controlled conditions, they cannot reliably manage real-time threat assessment, passenger safety decisions, or the dynamic personal protection component that defines this task. Current AI cannot perform the full end-to-end task of safely escorting individuals while protecting them from potential threats.
Task automatabilityclaude-sonnet-51/5This requires physical presence, driving, situational awareness, and the ability to physically intervene for protection—none of which current AI systems can perform end-to-end.
Adoption barriersclaude-haiku-4-5-202510015/5Personal protection and armed escort are heavily regulated; many jurisdictions legally require a licensed, bonded human security guard to perform or sign off on protective duties. Liability, criminal responsibility, and duty-of-care requirements create hard legal barriers to full automation.
Adoption barriersclaude-sonnet-54/5Personal protection often involves licensing (armed guard permits), liability for physical harm, and near-universal expectation of human presence for safety and trust.
Cost vs. human wageclaude-haiku-4-5-202510011/5Autonomous vehicle fleets remain expensive to deploy and maintain, and the liability and insurance costs for protective escort services without human security presence are prohibitively high. A security guard's loaded cost is substantially lower than the infrastructure and safety-overhead costs of AI-driven protection services.
Cost vs. human wageclaude-sonnet-51/5There is no viable AI substitute performing this physical task, so AI cost is not comparable—human labor remains the only option.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product reliably performs armed or unarmed personal protection escort autonomously today. Autonomous vehicles exist for transportation only, but lack the situational awareness, threat detection, and protective decision-making required for this security-specific task.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs personal protection escort or transport of individuals; autonomous vehicles remain limited and lack any protective/security function.

Respond to medical emergencies by administering basic first aid or by obtaining assistance from paramedics.

0

CI 0–0 · exposure 0 · augmentation 25 · importance 4.8/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Adoption of AI for actual emergency response is negligible because the task requires physical presence, real-time human judgment, and legal accountability that cannot be transferred to machines. Security roles remain heavily human-dependent.
Sector adoption velocityclaude-sonnet-51/5Security guard work is physical, low-digitization, and this specific emergency-response function shows no AI adoption trend in the field.
Augmentation potentialclaude-haiku-4-5-202510012/5AI could marginally assist by providing real-time first-aid guidance (e.g., via a voice agent) or by alerting paramedics automatically, but the human guard remains the active responder. Assistance is limited to information and coordination, not material productivity multiplier.
Augmentation potentialclaude-sonnet-52/5AI could assist indirectly via emergency dispatch apps, communication tools, or guidance scripts, but offers minimal enhancement to the core physical first-aid response itself.
Task automatabilityclaude-haiku-4-5-202510011/5AI systems cannot physically administer first aid, perform physical assessments, or summon paramedics in real emergencies. While AI might support decision-making in a purely informational role, the task inherently requires embodied human action and real-time judgment in high-stakes scenarios.
Task automatabilityclaude-sonnet-51/5Physical emergency response requiring hands-on first aid and human judgment in real time cannot be performed by AI systems; no robotic or software system can administer physical care today.
Adoption barriersclaude-haiku-4-5-202510015/5Strong legal and regulatory barriers exist: only certified humans may administer first aid, and liability for medical intervention falls on the person acting. Organizations face legal requirements that a trained human remain responsible for emergency response.
Adoption barriersclaude-sonnet-55/5Emergency medical response often involves duty-of-care, liability, and sometimes certification requirements (e.g., CPR/first-aid certification), and physical presence is legally and practically required.
Cost vs. human wageclaude-haiku-4-5-202510011/5AI has no meaningful cost advantage for this task because it cannot perform the core physical and judgment-heavy work. The human security guard remains mandatory; AI cannot substitute for their presence and actions.
Cost vs. human wageclaude-sonnet-51/5AI cannot substitute for the physical presence and action required, so there is no viable AI cost comparison—human labor is the only option.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed product can autonomously respond to medical emergencies or administer first aid. AI chatbots can provide information, but they cannot replace the human responder who must assess, act, and coordinate emergency services on-site.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product performs physical first aid or emergency medical response; this remains entirely outside current product capabilities.

Warn persons of rule infractions or violations, and apprehend or evict violators from premises, using force when necessary.

0

CI 0–0 · exposure 0 · augmentation 38 · importance 4.6/5 · click for rater detail

Panel ratings per dimension for this task
DimensionModelRatingRationale
Sector adoption velocityclaude-haiku-4-5-202510011/5Security remains a heavily human-dependent, physically-grounded sector with minimal automation of core apprehension and eviction tasks; adoption of autonomous enforcement is legally and organizationally blocked.
Sector adoption velocityclaude-sonnet-51/5Security guard work is physical, low-digitization, and shows minimal AI-driven displacement in current adoption data.
Augmentation potentialclaude-haiku-4-5-202510013/5AI-powered monitoring systems (cameras with anomaly detection, alerting) can assist human guards in identifying violations and situational awareness, moderately improving their ability to respond to infractions.
Augmentation potentialclaude-sonnet-52/5AI-enabled surveillance and detection systems can alert guards to potential infractions, offering some situational awareness support, but the physical confrontation and enforcement itself is unassisted by AI.
Task automatabilityclaude-haiku-4-5-202510011/5This task requires real-time physical presence, situational judgment, interpersonal communication, and use of force—none of which current AI systems can perform end-to-end. AI cannot apprehend, evict, or physically intervene.
Task automatabilityclaude-sonnet-51/5This task requires physical presence, real-time judgment, and potentially physical force to detain or evict individuals—capabilities entirely outside current AI systems' scope.
Adoption barriersclaude-haiku-4-5-202510015/5Hard legal and liability barriers exist: use of force is legally restricted to trained, licensed personnel; premises liability law requires human accountability; regulatory frameworks mandate human security personnel for apprehension and eviction.
Adoption barriersclaude-sonnet-55/5Use of force, detention, and eviction carry significant legal liability and often require licensed, authorized personnel, making this a hard-barrier task reserved for humans.
Cost vs. human wageclaude-haiku-4-5-202510011/5Replacing a security guard with current technology would require robotics, legal infrastructure, and continuous oversight that far exceeds the loaded wage of a human guard—cost-prohibitive today.
Cost vs. human wageclaude-sonnet-51/5There is no AI substitute capable of performing physical apprehension, so the comparison to human cost is not applicable and AI cannot deliver this output at any price.
Technical feasibility todayclaude-haiku-4-5-202510011/5No deployed AI product can legally or physically perform apprehension, eviction, or force application. Deployed systems may assist with monitoring or alerting, but cannot execute the core task.
Technical feasibility todayclaude-sonnet-51/5No deployed AI product can physically confront, apprehend, or evict a person; this remains purely a human physical-security function.

Related occupations — Protective Service

How to read this

A high substitution score does not mean this job disappears — it means a large share of its current tasks face replacement pressure, so the mix of tasks is likely to change. High augmentation alongside substitution typically means the occupation reorganizes around the protected tasks. Wide confidence intervals mean the rater panel disagreed: treat those scores as open questions, not verdicts.

What would change this score

New model capabilities (automatability, feasibility), falling inference costs (cost ratio), regulation and licensing shifts (barriers), and measured sector adoption (velocity) all re-enter at every index release. Each release is recomputed, versioned and kept queryable — scores are claims with a date on them, not permanent labels.